We have decided that leaking memory is a better fate than random,
non-debuggable memory corruption. Future CLs will begin changing
various fields to std::unique_ptr and std::shared_ptr.
It turns out that disabling the GC does not have disasterous impact.
The Nix evaluator only runs on the client CLI, never in any long-
running process. Even the REPL does not leak too badly under this
change, because it uses one EvalState for the duration of the REPL.
Building an explicitly tracing garbage collector is likely in the
future of this project, but that giant amount of work cannot be
done under a nix evaluator that is constantly crashing. We need to
restore development velocity here, and this is the best way we've
figured out to do it.
Change-Id: I2fcda8fcee853c15a9a5e22eca7c5a784bc2bf76
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1720
Reviewed-by: glittershark <grfn@gws.fyi>
Reviewed-by: tazjin <mail@tazj.in>
Tested-by: BuildkiteCI
The parser contained vectors, and the primary parser state, that
were not participating in GC tracing.
Change-Id: Ie198592cd7acffd390e3e2ae9595138b56416838
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1706
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
Reviewed-by: glittershark <grfn@gws.fyi>
Invocations of the MakeError macro that were not followed by a
semicolon messed up indentation in the next lines.
Change-Id: I03d7d1443f062a38af2c7da3da8928e0ed05e274
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1708
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Aded a few test cases covering the scanForReferences function, which had
been accidentally broken in 976a36c (which is now partially-reverted).
As part of this, since the test needed to generate hashes for store
paths, the logic in MakeStorePath to compress a sha256 hash down to 20
bytes and convert it to base32 has been extracted to a member function
on the Hash class.
Fixes: #34
Change-Id: Ie2d914688a80f42d0234d351a7cc0714fd15709e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1698
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
All of the miscellanious structs that make up a nix Value should inherit
from gc, as they contain pointers to GC'ed things as members
Fixes: #42
Change-Id: I057d8c9f7dafbee7de7644ff152c9cba1aa6bc03
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1696
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Now that we've fully implemented the RPC-based store client, we can get
rid of the UDSRemoteStore, whose only use was connecting to the locally
running nix daemon. The RemoteStore still needs to be around to connect
to remote upstream nix stores over SSH.
Change-Id: I0699819803cbfe966b9a46786f2c927d8e4bf1a2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1693
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Implement all remaining RPC calls on the RpcSstore client, remove a few
stub methods we had added that weren't actually present in the old
RemoteStore implementation, and add one more RPC call for getBuildLog
that is present in the store API, but that we hadn't added as a
stub *or* to the proto.
Fixes: #29
Change-Id: Id827f51a393ece4bc7bbecaf38aee9eb4b329770
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1692
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
Factor out the shared utilities for interacting with protobufs/grpc from
libstore to a proto.hh header in libproto.
Change-Id: I1cb8d94867d5d4b63a9994be0b53f8f612eb8e3a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1691
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
To aid in debugging RPC call failures, pass the name of the function
being called to the SuccessOrThrow util funcion in the RpcStore
Change-Id: I523dacfab896b85a3dbe6050c07ee6bd5906fa44
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1690
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Wrap the BuildPaths and AddTextToStore RPC handlers in HandleExceptions.
These were missed in the original pass due to a merge.
Change-Id: Ie5be45e6098fba7a2b6b1c1be81578cb742c2880
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1689
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
std::function has a natural null we can't eliminate anyway, so this was
effectively std::optional<std::optional<non_nullable_function>>.
Change-Id: If99f139146021edb25d133dad7f0f6e125ef53df
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1688
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Reviewed-by: glittershark <grfn@gws.fyi>
Implement both the client and server sides of AddToStoreNar, using a
templated generalization of the sources and sinks we were using for
AddToStore on both ends.
Change-Id: I73d0ed34118c711b125851dff99a7518ced4af35
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1686
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
The behavior to return a list containing a single empty string when
provided an empty string is a behavior that absl inherited from legacy
code. However, the behavior expected by legacy code in Nix is the
behavior provided by the SkipEmpty option. Switch all calls to use
SkipEmpty, except for the call already using SkipWhitespace.
See also commit 26a59482d2, with the
partly-prophetic message: "there may be other places we need to
fix this as well."
Change-Id: I6e94856a12cfb1b7e4a3b4e221769ed446648861
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1687
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This container implementation is much faster than std::map. We have
stuck to an ordered container because it's unclear whether the
accesses of this field (of which there are *many*) are actually
ordering dependent.
Also includes an Arbitrary implementation for absl::btree_map (for any
K, V that are also Arbitrary).
Change-Id: I04f58ca0ce32b9ae1759313b01508b0e44bae793
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1683
Reviewed-by: glittershark <grfn@gws.fyi>
Tested-by: BuildkiteCI
Bumps both nixos-unstable and nixos-20.03 to today's versions, as per
status.nixos.org
Contains minor fixes to things that broke because of the update:
* tazjin/frog: hardware.u2f is a deprecated setting
* glittershark/system: modSha256 in Go modules is now vendorSha256
* glittershark/owothia: removed version constraint on relude
Change-Id: Ib3e9612b1b06ed547b90e4f8b0ffe5ed7fe0a5c4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1642
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This builds on edef's work with depot-scan by adding a dedicated
flag to the command. We piggyback on upstream's restricted-mode
implementation, the checkSourcePath function.
Change-Id: I52bb613549f40dbca1e8caa036635910c1a3d6d0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1654
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Implement the RPC client and server handlers for the FindRoots and
CollectGarbage RPC calls
Change-Id: Ifa5d582c6a33bd1e7661ac2fc860505ef404dad0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1656
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
The source and destination strings cannot be the same string - absl
will write to the destination in a streaming manner, causing the
source to become invalid.
Change-Id: I3578cf1f8789a51d85e0950f7987c398f0a00953
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1659
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
In certain circumstances, the decompression thread could race ahead of the downloader thread and process the same chunk twice. Clear the data buffer while the lock is held to prevent this kind of incident.
Change-Id: I19a84a0c5768d1228c6c18a7664a7b8893ef96de
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1658
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This makes the configuration work out-of-the-box using CMake in a Nix
shell, rather than having to pass the additional variable to CMake on
the command line.
Change-Id: I04d3cc4f2e5ecf47bf2ee459d5e48588b84ae4dd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1643
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Reviewed-by: glittershark <grfn@gws.fyi>
These tests are in preparation for factoring the Store away from libutil's Callback to absl::StatusOr.
They use the newly added MockBinaryCacheStore.
Updates: #25
Change-Id: I30c207589ec38254806ebc9a983f35668e353ae9
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1595
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This adds an implementation of BinaryCacheStore to be used by tests exercising both the logic inherent to BinaryCacheStore and more general Store tests.
A new library target, nixstoremock, is created to indicate that this file is intended only for use in tests and not in user-facing code.
Change-Id: Ib68f777238843a4f3a2303db8a69735fbc22d161
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1645
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Implement the AddToStore RPC client method, which uses an
AddToStorePathWriterSink (the dual of the AddToStorePathReaderSource on
the server side) to hook into the dumpPath machinery (which we should
refactor not to use sinks or sources, but not yet) and write dumped
paths as binary data over gRPC. With this commit and sandboxing
disabled, the following derivation builds cleanly:
derivation {
name = "test";
builder = ./build.sh;
system = "x86_64-linux";
}
where build.sh has chmod +x and contains:
#!/bin/sh
echo 1 > $out
Change-Id: I94cab86f0825a3a9993262a9807130645c13bf44
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1638
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
Implement RPC handlers and client methods for the AddWithRoot and
SyncWithGC calls. With this and sandboxing disabled, the following
derivation is now building successfully:
derivation {
name = "test";
builder = "/bin/sh";
args = [ "-c" "echo 1 > $out" ];
system = "x86_64-linux";
}
Co-authored-by: Vincent Ambo <mail@tazj.in>
Change-Id: I245dca0dfca882e2eae5d93364287a8b79fead71
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1637
Reviewed-by: tazjin <mail@tazj.in>
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
Without this, these errors print as "uncaught exception of type
std::string", but with this they print using nix's built-in error
printer, which will be nice to have as we bang out the remaining store
API methods.
Co-authored-by: Vincent Ambo <mail@tazj.in>
Change-Id: I9bd4f77f65b27c8f97b727b98ef62e2061999e71
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1636
Reviewed-by: tazjin <mail@tazj.in>
Tested-by: BuildkiteCI
This reverts part of commit 976a36c2e4,
which was causing scanForReferences to hang indefinitely. I'm not sure
what the original intent of the commit was since there's not really much
of a commit message on it, but with this I am able to fully realise
derivations.
Change-Id: I620e892e2fe2244f066512286b086b3dd8f8f6c0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1635
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
Reviewed-by: kanepyork <rikingcoding@gmail.com>
This executable doesn't appear to ever actually look at argv, so
throwing an error if argc is wrong seems kinda silly.
Change-Id: Iabc9dad785d4c99be9e8fd4ecd954d38b876cecd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1633
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
point the SANDBOX_SHELL macro at the actual path to busybox on the build
machine, or allow it to be configured at build-time with a cmake option.
Change-Id: I044a1315ba9baa3bc9ceddf29f36d14f9f9ccd96
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1632
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
When tokenizeString was changed to absl::StrSplit, there was a behavior
change because tokenizeString on an empty string returned an empty
vector - which the derivation builder (and likely a bunch of other
stuff) was depending on. The canonical way of fixing this is by passing
absl::SkipEmpty() to the function - there may be other places we need to
fix this as well.
This commit also includes some opportunistic absl::StrFormats and
StrCats, because I was here anyway, but those have no semantic
difference.
Change-Id: Ibf9bb602284f793fa55728481f63b838fb7a41db
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1631
Reviewed-by: tazjin <mail@tazj.in>
Tested-by: BuildkiteCI
We wrap every server-side proto handler with a macro that catches
exceptions and turns them into proper grpc error codes. For the
time being, most exceptions map to INTERNAL, the existing mapping.
Change-Id: Id6ed6a279b198ad185d32562f39000ccc15eadbf
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1599
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
At some point the behavior of this function got changed as part of our
cleanup - this fixes it to behave the way the rest of the codebase
expects (and how it is documented in the header) and covers it with a
few tests.
Change-Id: Id4c91232968e73489cd866fb4a2a84bcf20d875e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1629
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
Two exceptions: IsValidPath needs to return success for invalid paths, and QueryAllValidPaths shouldn't need to check the paths it gets from itself.
Change-Id: I4d9d4125d34e8de42f30070aec607f8a902eded7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1598
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Skips over all the monobinary stuff and moves to a separate binary for
nix-daemon.
This also replaces the flag parsing logic with absl::flags. This
causes a behaviour change for --help, which no longer tries to display
a man page but instead shows the actual command-line help.
Note: This binary no longer links to the Boehm GC.
Change-Id: Ib852e994b82f2d56e91262878c10650e656427a9
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1622
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
The filename this is squatting is too good.
Change-Id: I694132b94d24786df84b6b5a698679bdce818b2e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1621
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Implement AddTextToStore and BuildPaths both on the client and the
server
Refs: #29
Change-Id: I45294c3e1c1a7489e42099d36425b7acc04e0427
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1560
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
This makes it easy to quickly run clang-tidy on tvix without seeing errors from the generated files.
Change-Id: I0e25089c5626aebdb5c016629a68da9ccd26c124
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1556
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Only libexpr depends on the garbage collector, specifically only
instantiations of EvalState actually require the GC to be initialised.
Rather than always starting it for the whole program, even if it is
not needed, this change moves the GC initialisation into libexpr,
guarded by absl::call_once.
This should make it possible to run the nix daemon without the garbage
collector interfering, granted that things are correcty separated and
the daemon does not actually invoke the evaluator.
Based on my investigation so far, the daemon logic itself does not
require libexpr to be present at all - so I think it is safe - but the
current monobinary might have some tricks up its sleeve that will
cause problems for us. We can deal with those if they arise.
Relates to https://b.tvl.fyi/issues/30
Change-Id: I61c745f96420c02e089bd3c362ac3ccb117d3073
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1584
Tested-by: BuildkiteCI
Reviewed-by: edef <edef@edef.eu>
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Add a rapidcheck test covering roundtrip parse and serialize for Nix
derivations. This covers a bug we discovered in ef54f5d which broke this
roundtrip.
Change-Id: I72d140334b5f24f79e82e34f98609c695dbfbf93
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1582
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
It turns out these aren't supposed to be reused after the first request
- reusing it was causing things to blow up.
Change-Id: I935f1ee82ed4e7f26528ae105c8fa2501fa81b97
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1559
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
Prepend the unix:// URI scheme to the daemon socket so that grpc knows
we want to connect to a unix socket rather than another type of URI. As
part of debugging this I made the failure message for the RPCStore
include the URI, which I'm leaving in since it'll be nice to have.
Change-Id: I6e70596895117b9a0d53fe2a61d8542ceb64c940
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1543
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Tested-by: BuildkiteCI
